There are some really big open questions out there about the role of human ingenuity in a world where AI is about to outpace us in an ever-growing variety of cognitive tasks.
I, for one, don't fear that humans are going obsolete. It's reassuring to remember that various forms of automation have been employed since the dawn of civilization, and a hyper-capitalist nation like the United States with every capability to employ said automations still has <4% unemployment.
A shoe factory that employed 100 line workers in 1924 may now, in 2024, consist of 3 robot technicians, 5 accountants, and 92 marketing managers; but if this is so it's because this is a more productive use of resources than the old arrangement.
What's different about AI and LLMs is that they seem to uniquely take aim at what David Graeber calls bullshit jobs — "professional, managerial, clerical, sales, and service workers." Graeber is a little too harsh when he refers to these jobs as "pointless." They do have a point, and they do provide economic value, otherwise the market would have found a way to avoid paying them. What makes them "bullshit" is that they are worth doing only in the context of the larger scaffolding of an overwrought and inefficient system.
Think about an administrator or claims adjuster at an enormous health insurance company. In the grand scheme of things we'd all probably experience better outcomes per dollar of healthcare spend if health insurance were nationalized and this role was obviated; but in the world of real-life America, these job continue to be performed because they provide a return on investment to someone, somewhere.
A second reason these jobs are "bullshit" is because they do not fully utilize our inherent human capacities for creativity, connection, and autonomy. Overall, AI is excellent at automating these sorts of low-to-mid-tier cognitive roles, which means that the people in them are going to have to move to areas where they hold competitive advantage over the bots.
We can speculate about what will be rare and valuable in this new world by starting with what will be cheap and abundant: Readable text. Whether it's the answers to customer service questions, medical or nutritional advice, itineraries for a weekend getaway, or relevant examples from case law, the ability to generate helpful text for some business purpose is about to drop to zero. That means we're about to have a massive surplus of words about things that you may or may not care about. Another phrase you might use for what I described in the previous sentence is "spam."
Over the last 3 months the quantity of cold email pitches in my inbox for startups or business software has absolutely exploded. And the quality and targeting of these messages has gone down — though I would imagine not so far down that the messages weren't worth it for the originator to send. Firing off 10,000 LLM-powered custom-seeming emails is way better for the sender than taking the time to write 100 legitimately custom emails, even if the effectiveness of each is 1/10th or 1/25th as good.
I used to reply to almost every cold email I received. Now my impulse is almost always to do the opposite, because my doubt that there's actually a person on the other end of the message is skyrocketing (as validated by the fact that the percentage of my responses that get any kind of response is going way down.) The more this happens, the more I crave an actual warm intro from somebody I trust and care about.
Venture capital was supposed to be an industry where social media and the open Internet was going to break down the "old boys’ club," where you didn't need to live in Palo Alto and know the right people to get your innovative idea funded. Yet now it seems like having pre-existing relationships with, or physical proximity to, the purveyors of private capital has never been more important. Is it a coincidence that this is the particular moment in time when Techstars is retreating from its second-tier metros? Isn't this precisely the opposite of what the Internet was supposed to do for us?
People want to believe in a world where your Twitter following is more important than where you want to college. Yet I can't help but notice that porn bots are making up a skyrocketing percentage of my newest followers. It seems like the best way to be Twitter famous in 2024 is to have built up a huge following on the platform in 2014. The more that the Internet fills up with spam, and the less it feels that there are genuine connections on the other side of the screen, the value of information will plummet and the value of legitimate human connection will go up.
This is what I mean by the title of this post. It's not the size of your network that will matter (because it has now become trivially easy to make superficial connections at scale.) It's about the depth of relationships and what can be accomplished in a network of trust. The aphorism "it's better to have 5 great friends than 500 decent friends" was briefly untrue from 2007-2020 but may now be truer than it's ever been.
As the price of synthetic connection falls, the value of genuine connection will rise. Yes, there will still be people on the margins who fall in love with their chatbot girlfriends or whatever. But in general I believe that most people will end up moderately repulsed by the blah-blah-blah that comes from our lifeless AI tools. And the role of the human being in the modern economy will be to offer something that the AI cannot: A real relationship with a real person.
Loved this. It will be interesting to see if “readable text” is the only media that becomes cheap and abundant or whether it expands into video.
Just last year I believed that sending cold emails with personalized Looms could stand out in my proprietary sourcing efforts, but now I’m not so sure.
Makes me think of the fake typing sound they play when you talk to a voice automated intake system when calling to schedule a dr appointment or calling an airline, or of how coffee shops (and cannabis dispensaries) differentiate themselves by having cute, charismatic baristas when their product could easily be sold via a vending machine.